ARTICLE

Bond Length and Charge Density

Variations within Extended Arm
Chair Defects in Graphene
Jamie H. Warner,†,* Gun-Do Lee,‡,* Kuang He,† Alex. W. Robertson,† Euijoon Yoon,‡ and Angus I. Kirkland†
†
Department of Materials, University of Oxford, Parks Road, Oxford OX1 3PH, United Kingdom and ‡Department of Material Science and Engineering,
Seoul National University, Seoul 151-744, Korea

ABSTRACT Extended linear arm chair defects are intentionally fabricated in

suspended monolayer graphene using controlled focused electron beam irradiation.
The atomic structure is accurately determined using aberration-corrected transmission
electron microscopy with monochromation of the electron source to achieve ∼80 pm
spatial resolution at an accelerating voltage of 80 kV. We show that the introduction of
atomic vacancies in graphene disrupts the uniformity of C C bond lengths immediately
surrounding linear arm chair defects in graphene. The measured changes in C C bond
lengths are related to density functional theory (DFT) calculations of charge density variation and corresponding DFT calculated structural models. We show
good correlation between the DFT predicted localized charge depletion and structural models with HRTEM measured bond elongation within the carbon
tetragon structure of graphene. Further evidence of bond elongation within graphene defects is obtained from imaging a pair of 5-8-5 divacancies.

KEYWORDS: graphene . HRTEM . electron microscopy . defects . TEM

T
he bonding strength between carbon
atoms is related to the charge density,
and in simple molecules, this can be
described by 'bond-order', with single, dou-
ble, triple bonds, and also aromaticity arising
from cyclic delocalization and resonances.
Graphene sheets exhibit long-range aromati-
city, equi-distribution of charge and hence
uniform C C bond lengths. A nanosized
fragment of graphene can be considered as
a polycyclic aromatic hydrocarbon (PAH), in
which the Pauling bond order is related to the
number of Kekulé resonance structures.1 3
Since most PAH or fullerenes are small in size,
there are a relatively small number of reso-
nance structures which lead to bond length
differences as recently confirmed by atomic
force microscopy.4 As the size of a PAH
increases, the number of Kekulé resonance
structures increases, and at the transition from
PAH (nanoscale) to graphene (microscale),
the number of these becomes sufficiently
large that all C C bond lengths are predicted
to be identical (∼0.142 nm).1
However, atoms at the edge of graphene
may have different C C bond values and
coordination geometries from those in the
bulk due to missing nearest neighbor atoms
and thus unique bonding arrangements.5 8
The introduction of defects into a pristine
graphene lattice breaks the lattice symme-
try locally.9 In particular, removal of atoms
from graphene initially forms vacancies,
which have been shown to subsequently
undergo reconfiguration in order to main-
tain local bonding to three nearest neighbor
atoms.9 11 However, this process is likely to
result in changes to the bonding charge
distribution and to the bond lengths for
C C atoms surrounding the defect structure.
The ability to detect small deviations in
C C bond lengths surrounding defect struc-
tures in graphene requires a local imag-
ing technique with sufficient resolution to
locate the position of individual carbon
atoms with suitable precision. Aberration-
corrected high resolution transmission elec-
tron microscopy (AC-HRTEM) is able to re- * Address correspondence to
solve the lattice structure in graphene.12 14 jamie.warner@materials.ox.ac.uk,
gdlee@snu.ac.kr.
However, in order to accurately measure the
position of atoms within graphene defects, Received for review July 10, 2013
the resolution limiting effects of damping and accepted October 15, 2013.
due to partial temporal coherence at low
Published online
accelerating voltages (80 kV) need to mini- 10.1021/nn403517m
mized. This can be achieved, either by elec-
tron optical correction of the chromatic C XXXX American Chemical Society

WARNER ET AL. VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX A

www.acsnano.org

aberration coefficient of the objective lens or by
monochromation of the electron source to reduce
the energy width.15 19
In this report, we demonstrate that a combination of
monochromated AC-HRTEM at low-voltage, multislice
image simulations and density functional theory (DFT)
calculations, provide insight into the mechanisms by
which defects in graphene directly affect the local
electronic charge distribution and bonding. To under-
stand the relationship between the charge distribution
predicted from DFT calculations and C C bond lengths
measured using AC-HRTEM, the defect structure studied
must be planar, normal to the electron beam, as out-of-
plane distortions skew the apparent C C bond length in
projection. Previous work has detected bond length
variations around dislocation cores in graphene, but as
dislocations are known to induce appreciable out-of-
plane distortions this limited conclusions relating the
charge to bond length.18,20 Linear defect structures
formed from the removal of atoms along the arm chair
direction, rather than along the zig-zag direction that
leads to dislocation pairs, are however planar. We there-
fore focus our present investigations on this type of
defect structure. Figure 1 shows a series of schematic
atomic models illustrating the missing atoms required to
form an extended linear arm chair defect (LAD) that
includes a four member carbon ring (tetragon). The
actual step-by-step formation process of this LAD is likely
to occur by the coalescence of smaller mono or divacancy
defects through bond rotations.
RESULTS AND DISCUSSION
For the studies reported here, monolayer graphene
was produced using chemical vapor deposition on a
copper substrate and transferred onto silicon nitride
TEM grids with 2 micron holes as previous reported.21
Controlled electron beam irradiation at 80 kV was used
to sputter carbon atoms from pristine regions of the as
prepared graphene to create the LAD structure using
conditions previously reported [9]. It should be noted
that other forms of defects were also produced within
the 10 nm area exposed to the focused electron beam.
Figure 2 presents a comparison of DFT relaxed atomic
models, multislice image simulations and strain maps
calculated using geometric phase analysis (GPA).22 24
The DFT model shows that the region of the LAD can be
considered planar, while the regions to the right that
contain a dislocation pair give rise to local out-of-plane
distortions. The image simulation closely matches the
experimental image of the LAD and the GPA shows
that there is significant strain arising from bond elon-
gation through the central region of the LAD.
A DFT calculated charge density map for the LAD
structure is shown in Figures 3a,b. The vertical bonds of
the carbon tetragons and terminating pentagonal
rings both show appreciable charge depletion. The
C C bond lengths calculated using DFT, Figure 3c,
WARNER ET AL.
ARTICLE
Figure 1. Schematic atomic models showing the formation
of the 5-8-4-8-4-8-5 linear arm chair defect structure by
atom removal. (a) Pristine graphene lattice (orange indi-
cates atoms subsequently removed). (b) Removal of three
carbon dimers (six atoms in total, orange atoms in (a)), along
the arm chair direction. (c) New bonds formed (yellow). (d)
After strain relaxation to form a 5-8-4-8-4-8-5 defect, with
carbon tetragons shown in blue.
show clear correlation between regions of charge
depletion and bond elongation, with the six vertical
C C bonds in the pentagonal rings and tetragons all
elongated between 5 and 10%. Changes in bond
lengths relative to a pristine region were calculated
and are expressed in color on the atomic model in
Figure 3c using the scale bar to the left.
To relate the bond elongations calculated using DFT
to our experimental data, we have compared an image
simulation of the LAD (Figure 4a) using the DFT model
(Figure 3c) to the experimental HRTEM image of the
LAD (Figure 4b). The C C bond lengths shown in
Figure 4a,b were determined using boxed line profile
scans across two atoms to measure the distance
between the maximum intensity in the peak positions,
with an uncertainty of ∼5 pm (see Figure 5d). There-
fore, measured bond lengths that exceed 153 pm are
beyond two standard deviations and are designated as
being 'above the noise' and correlate to real elonga-
tions of at least 1.07. A reference unstrained C C bond
length was measured at a distant location from the
LAD structure. The strain in each bond was sub-
sequently calculated by dividing each measured
bond distance in the LAD by this reference length,
(>1, elongation and <1, compression). Measurements
from the image simulation (Figure 4d) and experimen-
tal image (Figure 4e) demonstrate that the bond
elongation in the LAD predicted from DFT calculations
can be detected from HRTEM images. Importantly, all
three measurements in Figure 4c e show that at least
5 of the 6 vertical bonds in the central region of the LAD
are substantially elongated between 5 and 10%. Some
of the other bonds, such as on the right-hand-side of
the pentagon/octagon, also show significant elonga-
tion, which we attribute to the influence of a nearby
dislocation pair that cause distortion to the lattice and
is hard to accurately simulate by DFT due to the large
number of atoms.
VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX B
www.acsnano.org
 ARTICLE
Figure 2. Comparisons of the experimental structure of a 5-8-4-8-4-8-5 linear arm chair defect (LAD) with a density functional
theory (DFT) model. (a) Top view of the DFT model of defect structures observed to be produced by electron beam sputtering
in graphene. (b) Side view of (a) showing out-of-plane distortions induced by dislocations, but with a relatively flat region
around the tetragons in the LAD. (c) 3D perspective view showing valleys and hills in the graphene lattice from dislocations.
(d) Multislice image simulation using the atomic model in (a) (red boxed region). (e) Experimental HRTEM image of a LAD
created in graphene by electron beam induced sputtering. (f) Cropped region of the experimental image of the LAD structure
at higher magnification. (g) Strain component (εyy) calculated using Geometric Phase Analysis (GPA) of the image in (f), color
bar indicates the scale used. (h) Overlay of the GPA analysis and HRTEM image showing the relative location of the lattice
strain with the atomic structure.

Figure 3. Correlation of charge density and bond length. (a) Total charge density determined using DFT for the atomic
structure shown in Figure 2a. The unit of charge density is e/a03 (a0 is Bohr radius). (b) Higher magnification map of the total
charge density for the 5-8-4-8-4-8-5 LAD structure showing charge depletion within the carbon tetragon. (c) Bond lengths in
angstroms, obtained from DFT calculations for the structure in Figure 2a, focused on the 5-8-4-8-4-8-5 LAD. Error in bond
length values (0.01 Å.

The finer structural details of the carbon tetragon are
compared in Figure 4f i. These show that in the tetra-
gon, bond elongation (measured from HRTEM images)
is related to the calculated charge depletion in the
vertical bond. Although the match between the all the
C C bond lengths in the DFT models, image simulations
and HRTEM images is not exact, a dominant trend
showing bond elongation within the central region of
the LAD structure is indicated in all cases and within the
experimental uncertainty of the measurements.

WARNER ET AL. VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX C

www.acsnano.org
 ARTICLE
Figure 4. Charge density dependent bond length variations in carbon tetragons. (a) Multislice image simulation, as in
Figure 2d, with a 'fire' color LUT for display. (b) HRTEM image, as in Figure 2e, of the 5-8-4-8-4-8-5 defect structure, with a 'fire'
color LUT display. (c) Ratio of local bond length to unperturbed value, R = Bl/Bu, based on atomic positions from DFT
calculation. (d) R value determined from the multislice image simulation in (a). (e) R value determined from the HRTEM image
in (b). (f) Atomic model based on DFT calculated atomic positions for the right carbon tetragon. (g) Total charge density, as in
Figure 3b of the right carbon tetragon. (h) Cropped region of (a) showing the right carbon tetragon and R values. (i) Cropped
region of (b) showing the right carbon tetragon and R values.

Figure 5. (a) Monochromated AC-TEM image of two 5-8-5 divacancy defects in graphene. (b) Atomic model of two 5-8-5
divacancy defects in graphene (8 atom ring = orange, 5 atom ring = red). (c) Higher magnification image of (a) with measured
bond lengths in picometers. (d) AC-TEM image of a region of pristine graphene with measured bond lengths in picometers
and an average C C distance of 143 ( 5 pm.

Further confirmation that bond length variations higher magnification TEM image of the two divacan-
exist within graphene defect structures was obtained cies is presented in Figure 5c and a region of pristine
by examining another defect composed of two neigh- graphene for comparison in Figure 5d. Measurement
boring 5-8-5 divacancies, as shown in Figure 5a,b. A of the distance between the centers of the contrast

WARNER ET AL. VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX D

www.acsnano.org

spots for nearest neighbors in the pristine graphene
area, Figure 5d, yields an average C C distance of
143 ( 5 pm. In Figure 5c, all four C C bonds at the
interface between the 5 8 rings show elongated
values of 167, 162, 170, and 153 pm. These values lie
beyond a standard deviation and are statistically sig-
nificant and represent real changes in individual C C
bond lengths. The total charge density of this defect
structure was calculated by DFT and is shown in Figure 6.
The four C C bonds at the interface between the 5 8
Figure 6. Total charge density of the 5-8-5 divacancy pair
calculated by DFT for the atomic structure shown in Figure 5b.
The unit of charge density is e/a03 (a0 is Bohr radius).
Figure 7. Ratio of local bond length relative to an unperturbed value for (top) DFT calculated structure and (bottom)
experimental measurements taken from Figure 5c.
WARNER ET AL.
ARTICLE
rings that were experimentally shown to have elongated
values are also the regions with the lowest charge density
in Figure 6.
Figure 7 shows a comparison between the DFT
predicted C C bond lengths and those measured
experimentally in Figure 5, depicted as a ratio of local
bond length to unperturbed value, as in Figure 4c e.
The four major elongated bonds in Figure 7a are
highlighted in pink. The experimental data also shows
three of these bonds as pink, with the remaining bond
partially elongated. These results confirm that C C
bonds within graphene that have lower charge density
are longer and our HRTEM is capable of imaging these
elongations with confidence above the noise level.
CONCLUSION
The polycrystalline continuous graphene sheets ex-
amined in this work are at least a millimeter in size, with
single crystal sizes of ∼100 200 μm.21 While we do not
directly measure charge density, we have measured
C C bond lengths and compared these to DFT pre-
dicted bond lengths. The reason the predicted bond
lengths by DFT are not all identical (i.e. 142 pm) is due
to the system being comprised of purely carbon and
VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX E
www.acsnano.org

the relevant distribution of charge density between
bonds. If there was no charge density variation, there
would be no C C bond length variation. Therefore, the
changes we experimentally measure by HRTEM can be
related to the DFT bond lengths and the DFT calculated
charge density changes. The removal of atoms from a
small central region leads to significant variations in
the predicted bond lengths surrounding the defect
formed and we have used DFT to understand how this
relates to charge density changes. The planar structure
of the tetragon in the LAD structure of graphene
METHODS
Density Functional Theory Calculations. We performed density
function theory calculation within the local density approxima-
tion by using a plane wave basis pseudopotential method as
implemented in Vienna ab initio simulation package code.25
Vanderbilt pseudopotentials are also used in this calculation.26
The basis set contains plane waves up to an energy cutoff of
400 eV. In these electronic structure calculations, we performed
two calculations. One is a calculation for a large system with
878 carbon atoms in the unit cell containing a 5-8-4-8-4-8-5
defect, a SW defect, and dislocation cores, as shown in Figure 2a.
The other is a calculation for a small system with 506 carbon
atoms in the unit cell containing only 5-8-4-8-4-8-5 defect to
study intensively the effect of 5-8-4-8-4-8-5 defect, as shown in
Figure 3a. The small system is large enough to neglect the interac-
tion between 5-8-4-8-4-8-5 defects in neighboring unit cells.18
Considering the missing carbon atoms from the pristine graphene
structure, we contracted the graphene lattice parameter along the
y-axis by 3.1% and 2.7% for the large system and small system,
respectively. The unit cells also contain a vacuum region of 15 Å.
Periodic boundary conditions are used in all three dimensions. The
Brillouin zone was sampled using a (1  2  1) Γ-centered mesh in
the large system and using (2  2  1) Γ-centered mesh in the
small system. When structural relaxations are performed, the
structure is relaxed until the force on each atom is smaller than
0.02 eV/Å.
Transmission Electron Microscopy. HRTEM was performed using
Oxford's JEOL JEM-2200MCO field-emission gun transmission
electron microscope, fitted with CEOS probe and image aberra-
tion correctors and a double Wien Filter monochromator oper-
ated with a 5 μm slit to reduce the energy spread of the electron
beam to 217 meV at an accelerating voltage of 80 kV.18 Data was
recorded using a Gatan Ultrascan 4k  4k CCD camera with 1 2 s
acquisition times. The typical beam current density used for
imaging was measured using a faraday cup and estimated to
be ∼(0.2 2)  106 e/(nm2 3 s). HREM image simulations were
performed using JEMS software with supercells. Supercells were
constructed using Accelrys Discovery Studio Visualizer.
Image Processing and Analysis. HRTEM images were processed
using ImageJ. First a raw image (2k  2k pixels) is subjected to a
bandpass filter, between 100 and 1 pixels, in order to remove
the long-range uneven beam illumination and then smoothed.
Multiple frames (3 5) of the exact same structure are then
averaged to increase signal-to-noise. A color look-up-table
(LUT) of 'Fire' is used to improve visual inspection of images.
The bond lengths are measured by locating the center of
contrast spots in the HRTEM image and measuring the distance
between them. The contrast spots associated with atoms in
graphene are not exactly circular due to the hexagonal pattern,
but also the presence of defects causes deviations in the
symmetry of the hexagonal lattice. These effects complicate
the use of peak finding algorithms or Gaussian fits to line-
profiles for accurate bond length comparisons within defects.
Instead we used a combination of peak finding software with
manual correction/adjustment by visual inspection for locating
the best-fit for the center of the contrast spots. We have
WARNER ET AL.
ARTICLE
provided an excellent platform to demonstrate the
correlation between calculated charge density and
experimentally measured bond lengths. An analysis
of a pair of divacancies also revealed substantial bond
elongation at the 5 8 ring interface region with the
lowest charge density and corroborates the findings
on the extended LAD structure. This is the first direct
confirmation of such correlation at the single atom
level in a sample that is orders of magnitude larger
than small PAHs or fullerene molecules previously
studied.
previously shown this approach to be valid for mapping bond
lengths in sp2 hexagonal systems in ref 27.
Conflict of Interest: The authors declare no competing
financial interest.
Acknowledgment. J.H.W. thanks the Royal Society and Bal-
liol College for support. G.-D.L and E.Y. acknowledge support
from the National Institute of Supercomputing and Networking/
Korea Institute of Science and Technology Information
with supercomputing resources including technical support
(KSC-2013-C3-005) and support from the National Research Foun-
dation of Korea funded by the Ministry of Education, Science and
Technology under Basic Science Research (No. 2010-0012670),
BK21 and WCU program (R31-2008-000-10075-0).
Supporting Information Available: Simulations showing the
effect of astigmatism on the images of extended arm chair
defects in graphene and the measurement of a second LAD
structure. This material is available free of charge via the
Internet at http://pubs.acs.org.
REFERENCES AND NOTES
1. Pauling, L.; Brockway, L. O.; Beach, J. Y. The Dependence of
Interatomic Distance on Single Bond-Double Bond Reso-
nance. J. Am. Chem. Soc. 1935, 57, 2705–2709.
2. Randic, M. Aromaticity of Polycyclic Conjugated Hydro-
carbons. Chem. Rev. 2003, 103, 3449–3605.
3. Sedlar, J.; Andelic, I.; Gutman, I.; Vukicevic, D.; Graovac, A.
Vindicating the Pauling-Bond-Order Concept. Chem. Phys.
Lett. 2006, 427, 418–420.
4. Gross, L.; Mohn, F.; Moll, N.; Schuler, B.; Criado, A.; Guitian,
E.; Pena, D.; Gourdon, A.; Meyer, G. Bond-Order Discrimina-
tion by Atomic Force Microscopy. Science 2012, 337,
1326–1329.
5. Koskinen, P.; Malola, S.; Hakkinen, H. Self-Passivating Edge
Reconstructions of Graphene. Phys. Rev. Lett. 2008, 101,
115502.
6. Suenaga, K.; Koshino, M. Atom-by-Atom Spectroscopy at
Graphene Edge. Nature 2010, 468, 1088.
7. Koskinen, P.; Malola, S.; Hakkinen, H. Evidence for Gra-
phene Edges Beyond Zigzag and Armchair. Phys. Rev. B
2009, 80, 073401.
8. Chuvilin, A.; Meyer, J. C.; Algara-Siller, G.; Kaiser, U. From
Graphene Constrictions to Single Carbon Chains. New J.
Phys. 2009, 11, 083019.
9. Robertson, A. W.; Allen, C. S.; Wu, Y. A.; He, K.; Olivier, J.;
Neethling, J.; Kirkland, A. I.; Warner, J. H. Spatial Control of
Defect Creation in Graphene at the Nanoscale. Nat. Com-
mun. 2012, 3, 1144.
10. Kotakoski, J.; Krasheninnikov, A. V.; Kaiser, U.; Meyer, J. C.
From Point Defects in Graphene to Two-Dimensional
Amorphous Carbon. Phys. Rev. Lett. 2011, 106, 105505.
11. Banhart, F.; Kotakoski, J.; Krasheninnikov, A. V. Structural
Defects in Graphene. ACS Nano 2011, 5, 26–41.
12. Meyer, J. C.; Kisielowski, C.; Erni, R.; Rossell, M. D.; Crommie,
M. F.; Zettl, A. Direct Imaging of Lattice Atoms and
VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX F
www.acsnano.org
 ARTICLE
Topological Defects in Graphene Membranes. Nano Lett.
2008, 8, 3582–3586.
13. Hashimoto, A.; Suenaga, K.; Gloter, A.; Urita, K.; Iijima, S.
Direct Evidence for Atomic Defects in Graphene Layers.
Nature 2004, 430, 870–873.
14. Girit, O.; Meyer, J. C.; Erni, R.; Rossell, M. D.; Kisielowski, C.;
Yang, L.; Park, C-H; Crommie, M. F.; Cohen, M. L.; Louie, S. G.;
et al. Graphene at the Edge: Stability and Dynamics.
Science 2009, 323, 1705–1708.
15. Kabius, B.; Hartel, P.; Haider, M.; Muller, H.; Uhlemann, S.;
Loebau, U.; Zach, J.; Rose, H. First Application of
Cc-Corrected Imaging for High-Resolution and Energy
Filtered TEM. J. Electron Microsc. 2009, 58, 147–155.
16. Haider, M.; Hartel, P.; Muller, H.; Uhlemann, S.; Zach, J.
Information Transfer in a TEM Corrected for Spherical and
Chromatic Abberation. Microsc. Microanal. 2010, 16, 393–
408.
17. Mukai, M.; Kaneyama, T.; Tomita, T.; Tsuno, K.; Terauchi, M.;
Tsuda, K.; Naruse, M.; Honda, T.; Tanaka, M. Performance of
a New Monochromator for a 200 kV Analytical Electron
Microscope. Microsc. Microanal. 2005, 11, 2134.
18. Warner, J. H.; Margine, E. R.; Mukai, M.; Robertson, A. W.;
Giustino, F.; Kirkland, A. I. Dislocation Driven Deformations
in Graphene. Science 2012, 337, 209–212.
19. Warner, J. H.; Mukai, M.; Kirkland, A. I. Atomic Structure of
ABC Rhombohedral Stacked Trilayer Graphene. ACS Nano
2012, 6, 5680–5686.
20. Liu, Y.; Yakobson, B. I. Cones, Pringles, and Grain Boundary
Landscapes in Graphene Topology. Nano Lett. 2010, 10,
2178–2183.
21. Wu, Y. A.; Fan, Y.; Speller, S.; Creeth, G.; Sadowski, J.; He, K.;
Robertson, A.; Allen, C.; Warner, J. H. Large Single Crystals
of Graphene on Melted Copper Using Chemical Vapour
Deposition. ACS Nano 2012, 6, 5010–5017.
22. Cowley, J. M.; Moodie, A. F. The Scattering of Electrons by
Atoms and Crystals. I. A New Theoretical Approach. Acta
Crystallogr. 1957, 10, 609–619.
23. Goodman, P.; Moodie, A. F. Numerical Evaluations of
N-Beam Wave Functions in Electron Scattering by the
Multi-Slice Method. Acta Crystallogr., A. 1974, 30, 280–290.
24. Hytch, M. J.; Snoeck, E.; Kilaas, R. Quantitative Measure-
ment of Displacement and Strain Fields from HREM
Micrographs. Ultramicroscopy 1998, 74, 131–146.
25. Kresse, G.; Furthmuller, J. Efficient Iterative Schemes for
Ab Initio Total-Energy Calculations Using a Plane-Wave
Basis Set. Phys. Rev. B 1996, 54, 11169–11186.
26. Vanderbilt, D. Soft Self-Consistent Pseudopotentials in a
Generalized Eigenvalue Formalism. Phys. Rev. B 1990, 41,
7892–7895.
27. Warner, J. H.; Young, N. P.; Kirkland, A. I.; Briggs, G. A. D.
Resolving Strain in Carbon Nanotubes at the Atomic Level.
Nat. Mater. 2011, 10, 958–962.

WARNER ET AL. VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX G

www.acsnano.org